COMAD 2005
INVITED INDUSTRY SESSION
Talk 1
TESLA: On-demand Information Systems
V S Batra
(IBM India)
ABSTRACT:
The Tesla project aims to build or enhance middleware to give
application programs flexible and predictable access to a large numbers of autonomous and heterogeneous data sources and
compute resources. We seek to provide flexibility for data access through a
novel query formulation and query processing scheme that gives the abstraction
of a highly available and performant virtual data
store over logical data domains. This abstraction masks from application
programs events such as addition or removal of data sources, substitution of
replicas for performance and availability, and source unavailability, by
converting them into degradations (or improvements) in the information quality.
We also aim to provide predictable quality of service for access to this virtual
data store, without relying on the dedicated resources. Instead, Tesla uses a
variety of resource managers to dynamically provision compute resources as
needed, to meet the quality of service needs of the application workload.
SPEAKER BIO:
Vishal Batra is a 1997 graduate from
IIT Kharagpur. He joined IBM Research in 2001 and has
worked on different domains such as Life Sciences, Unstructured Information
Management and Grid Computing. Prior to joining IBM, he has worked on various
consulting projects and delivered eCommerce applications to clients such as
Barclays Bank,
Talk 2
Emerging Trends in OLAP
Vaishnavi Sashikanth (
ABSTRACT:
It’s
more than a decade since Arbor Software pioneered the field of OLAP with a
product called Essbase that brought multidimensional access to spreadsheet users.
The definition of a sound
multidimensional data model, effective dimensional-cube persistence mechanisms,
aggregations, ad-hoc analyses involving pivoting, rollups, drill downs, an
expressive declarative language in which to express multidimensional analysis
are just a handful of problems that still continues to intrigue academic researchers, relational and OLAP database vendors. However, the focus thus far has predominantly
been to efficiently support multidimensional reporting. In this talk we will present the emerging
needs of analytical applications, new data modeling requirements, the role of
unstructured data and the notion of closed-loop analysis and the technological
requirements it entails. We will close
by highlighting briefly some of the challenges we have tackled in the context of
Essbase.
SPEAKER BIO:
Vaishnavi (Anjur) Sashikanth
is the Vice President and Architect for the Essbase
suite of products at Hyperion Solutions. During her tenure with Hyperion over
the last 6 years, Vaishnavi rearchitected
core components of the server optimizing the engine to work with any of a
relational, multi-dimensional or hybrid physical representation of an OLAP
cube. Under her leadership, Essbase server evolved to
support two different cube storage and aggregation mechanisms each optimized
for different data distribution patterns, transparently, within one engine -
setting a new architectural standard for Multidimensional OLAP engines. Prior
to Hyperion, Vaishnavi worked at Sybase on revamping
the storage access and transaction management components of Sybase Adaptive
database server Vaishnavi holds a Masters degree in
Computer Science from the University of Wisconsin, Madison. She received her
Bachelors degree, also in Computer Science from College of Engineering, Guindy in India.
Talk 3
Mobile Data Analysis and Reporting
Prasad Ram (Yahoo India)
ABSTRACT:
Corporations are increasingly offering newer
services on a global scale that are consumed on handheld devices (such as PDAs, cell phones) by
MARS is a
highly scalable Mobile Data Analysis and Reporting System developed at Yahoo
that offers a global platform for mobile transaction data collection, analysis,
tracking and reporting to global business units and partners. It collects and
synchronizes all transaction data across diverse international data sources,
organizes data in global data mart, and offers different views and reports to
various groups. MARS provides a uniform (and extensible) data model that
normalizes data across sources and mechanisms for efficiently mapping multiple data
formats. It also offers hierarchical (and extensible) namespaces for data
organization, end-to-end tracking of transactions and multi-level reporting. It
addresses varying needs on the timeliness of information batch to near
real-time in different contexts and ensures controlled access to data by
multiple groups/individuals.
SPEAKER BIO:
Dr. Prasad Ram is the Chief Technology
Officer of